US7463646B2 - Method and system for fibre channel arbitrated loop acceleration - Google Patents

Abstract

A fibre channel switch element with an acceleration device that connects plural fibre channel devices in an arbitrated loop and monitors frames sent during a loop initialization process is provided. The acceleration device creates an AL_PA table selected by each fibre channel device, wherein the AL_PA table is used for port selection; and during an arbitration process sends a benign primitive to non-arbitrating devices. The acceleration device includes, a global arbitration module, a state machine module, and a matrix for connecting plural fibre channel devices.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority under 35 USC Section 119(e), to the following provisional patent applications:

• • Ser. No. 60/487,876 filed on Jul. 16, 2003;
  • Ser. No. 60/487,887 filed on Jul. 16, 2003;
  • Ser. No. 60/487,875 filed on Jul. 16, 2003;
  • Ser. No. 60/490,747 filed on Jul. 29, 2003;
  • Ser. No. 60/487,667 filed on Jul. 16, 2003;
  • Ser. No. 60/487,665 filed on Jul. 16, 2003;
  • Ser. No. 60/492,346 filed on Aug. 4, 2003; and
  • Ser. No. 60/487,873 filed on Jul. 16, 2003.

BACKGROUND

The disclosures of the foregoing applications are incorporated herein by reference in their entirety.

1. Field of the Invention

The present invention relates to networks, and more particularly, to reducing latency in a fibre channel arbitrated loop environment.

2. Background of the Invention

Fibre channel is a set of American National Standard Institute (ANSI) standards, which provides a serial transmission protocol for storage and network protocols such as HIPPI, SCSI, IP, ATM and others. Fibre channel provides an input/output interface to meet the requirements of both channel and network users.

Fibre channel supports three different topologies: point-to-point, arbitrated loop and fibre channel fabric. The point-to-point topology attaches two devices directly. The arbitrated loop topology attaches devices in a loop. The fibre channel fabric topology attaches host systems directly to a fabric, which are then connected to multiple devices. The fibre channel fabric topology allows several media types to be interconnected.

Fibre channel is a closed system that relies on multiple ports to exchange information on attributes and characteristics to determine if the ports can operate together. If the ports can work together, they define the criteria under which they communicate.

In fibre channel, a path is established between two nodes where the path's primary task is to transport data from one point to another at high speed with low latency, performing only simple error detection in hardware.

FC-AL is one fibre channel standard (incorporated herein by reference in its entirety) that establishes the protocols for an arbitrated loop topology.

In a Fibre Channel Arbitrated Loop (FC-AL) implementation of each device on the loop must have the intelligence to process incoming data in accordance with FC-AL rules. Latency occurs due to every device in the loop, as data is processed. This latency is accumulated around the loop. Even when a device is only required to pass data to the next device, it still has latency before passing on the data, because the device must continuously process data for changes in state. The accumulated latency slows down data transfers in a heavily populated loop.

The traditional FC-AL implementation has a transmitter of one device connected to the receiver of the next device and so on until the transmitter of the last device is connected to the receiver of the first device, creating a loop, as shown inFIG. 2.

In the traditional FC-Al implementation, all fibre channel devices (“FCDs”), to properly follow the FC-AL protocol, need to implement a LPSM. The LPSM has the following states: initializing, open-initiate, monitoring, arbitrating, arbitration won, open, opened, transmitted close, received close, transfer, and old-port (optional) before the FC-AL standard.

Under the FC-AL standard, there can be a maximum of 127 devices on a loop. Every device on the loop must be operating correctly for the loop to be operational. Because of this, many implementations include a bypass element that allows loop connectivity to be maintained when a particular device is not available for proper operation. Standard bypass elements are not intelligent devices and are not controlled based on data being transmitted but on device functionality or availability. These bypass elements do not add latency, but add jitter.

Each connected device must follow the FC-AL standard by decoding primitives, maintaining a Loop Port State Machine (LPSM), obtaining AL_PA's through the LIP process, performing arbitration, and sending frames in a manner that is consistent with the standard.

The following introduces some FC-AL process steps and how these processes add delay/latency:

LIP Process: In the traditional FC-AL implementation, the LIP process is started when one or more of the devices on the loop start sending LIPs to the next device in the loop. After all devices have seen the LIP, a temporary master is selected and this master sends out various frames that each device uses to select an AL_PA. The frames are sent in a serial fashion and the master must get the first frame back before sending the next type of frame. There are several trips around the loop before the LIP process is completed. Each device acquires an AL_PA and only needs to know its own AL_PA to be FC-AL compliant. Thus when all the frames return to the master, all the participating devices have their AL_PA's. The AL_PA's are used for identification and priority evaluation in the arbitration.

Each device in the loop takes time to process incoming data and deciding (per FC-AL standard) what to send out to its transmitter. Thus each device adds to latency and slows data processing. Two major points of latency are the time delay to create (and close) the loop circuit and the time delay for frames to be passed back and forth. To create a loop circuit in a traditional implementation the device must arbitrate for access and then open the device.

Arbitration process: In the traditional FC-AL implementation, the device desiring access to the loop does arbitration. In order to arbitrate and win arbitration, the device must send arbitration primitives and receive the arbitration primitive back.

In the simplest case, only one device wants access. The device that wants access to the loop sends its arbitration primitive to the next device in the loop and after some processing the next device passes the arbitration primitive onto the next device and so on until the primitive returns to the originating device. Therefore, for the FCD to know it has won arbitration there is an entire loop delay, even if it is the only device that wants access to the loop.

Open process: In the traditional FC-AL implementation to complete a loop circuit, the winning device needs to open the device it wants to communicate with. To open the other device and know that it is open (using zero BB_Credit) the device sends an open primitive and receives back an R_RDY primitive. The open primitive (OPN) is sent by the winning device to the next device, which processes it and sends it on to the next device and so on until the destination device receives the open primitive. Then the destination FCD sends a R_RDY primitive to the next device it is connected to and that device processes the primitive and sends it to the next device and so on, until the R_RDY primitive reaches the winning FCD. Thus to open the loop circuit and to know it is open, causes a delay around the loop.

“Open Replicate” process: In the traditional FC-AL implementation, to complete the loop circuit a winning FCD can send an open replicate primitive(s). When the winning FCD does a broadcast open replicate then every FCD will replicate the incoming frames and send them out to the next device. When the winning FCD does a specific open replicate then it sends out a number of open replicate primitives to inform the FCDs (that it wants to communicate) to replicate the frames on the incoming port and also send the frame on to the next device. There are no R_RDY primitives sent for an open replicate primitive, so frames can quickly follow the open primitive(s), but there is still delay to pass the frames around the entire loop.

Close process: In the traditional implementation to close the loop circuit, active devices must send a close primitive to the other device. An entire loop delay occurs when one of the devices sends a close primitive that is received by the other device before it sends its close primitive. Once the first device receives the second close primitive then the loop circuit is closed. Per the FC-AL standard, it is not required for the close primitives to be sent in a serial fashion, hence, the close primitives can be traveling the loop at the same time affecting the overall latency.

The following summarizes, the problems with traditional FC-AL implementation:

Problem 1: Large latency in loop communication implementations.

Problem 2: Separate devices to bypass non-working devices add cost to overall system and add jitter to the communication network.

Problem 3: Since only a device needs to know its AL_PA in a conventional standard system, there is no required “system” for determining the AL_PAs used in a loop.

Problem 4: Conventional methodology does not support global arbitration.

Problem 5: All devices on a loop currently need to support a full LPSM, and hence can be expensive.

Problem 6: The open replicate feature of the FC-AL has a long latency on frames sent because all frames need to return to the source device to be removed from the loop.

Therefore, what is required is a process and system that can accelerate frame processing and minimize latency in fibre channel arbitrated loop topology.

SUMMARY OF THE INVENTION

In one aspect of the present invention, a fibre channel switch element is provided. The switch element includes, an acceleration device that connects plural fibre channel devices in an arbitrated loop and monitors frames sent during a loop initialization process and creates an AL_PA table selected by each fibre channel device, wherein the AL_PA table is used for port selection; and during an arbitration process sends a benign primitive to non-arbitrating devices.

During an open process, the acceleration device connects a fibre channel device that wins arbitration to a destination device so that an open primitive can be sent directly to the destination device. Non-destination devices receive a benign primitive during the open process to maintain proper loop state machine state.

During an open replicate process, the acceleration device after it receives an open replicate primitive creates a spray pattern for all ports with which a fibre channel device intends to communicate.

In yet another aspect of the present invention, an acceleration device used by a fibre channel switch element is provided. The acceleration device includes, a global arbitration module, a state machine module, and a matrix for connecting plural fibre channel devices; wherein the acceleration device connects plural fibre channel devices in an arbitrated loop and monitors frames sent during a loop initialization process and creates an AL_PA table selected by each fibre channel device, wherein the AL_PA table is used for port selection; and during an arbitration process sends a benign primitive to non-arbitrating devices.

In yet another aspect of the present invention, a method for accelerating traffic in a fibre channel arbitrated loop topology having a fibre channel switch element with an acceleration device is provided. The method includes, monitoring fibre channel frames during a loop initialization process; creating an AL_PA table for fibre channel devices connected in the arbitrated loop and using the AL_PA table for identifying ports. The method also includes, sending benign primitives to non-arbitrating devices during an arbitration process; connecting a fibre channel device that sends an OPEN primitive to a destination port; and sending a benign primitive to non-destination device connected in the arbitrated loop.

The method also includes creating a spray pattern after receiving an open replicate primitive from a fibre channel device; and sending frames to plural destination ports at the same time.

This brief summary has been provided so that the nature of the invention may be understood quickly. A more complete understanding of the invention can be obtained by reference to the following detailed description of the preferred embodiments thereof concerning the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing features and other features of the present invention will now be described with reference to the drawings of a preferred embodiment. In the drawings, the same components have the same reference numerals. The illustrated embodiment is intended to illustrate, but not to limit the invention. The drawings include the following Figures:

FIG. 1 shows a block diagram of a storage area network;

FIGS. 2 and 3 show configurations that use the adaptive aspects of the present invention;

FIG. 4 shows a block diagram of a switch element, according to one aspect of the present invention;

FIGS. 5A and 5B (jointly referred to asFIG. 5) show a block diagram of a transmission protocol engine, according to one aspect of the present invention;

FIGS. 6A and 6B show block diagrams for a diagnostic module and a SES module, according to one aspect of the present invention;

FIG. 7 shows a block diagram of acceleration device, according to one aspect of the present invention;

FIG. 8 shows an example of how plural fibre channel switch elements may be used to couple 127 fibre channel devices, according to one aspect of the present invention; and

FIGS. 9A and 9B show a state machine for the acceleration device, according to one aspect of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Definitions:

The following definitions are provided as they are typically (but not exclusively) used in the fibre channel environment, implementing the various adaptive aspects of the present invention.

“Active device”: The source device (winning arbitrator) or the destination device.

“Active port”: A port with an active device attached to it.

“AL_PA”: Arbitrated loop physical address.

“BB_Credit”: Buffer-to-Buffer Credit

“FCD”: Fibre Channel device

“FC-AL”: Fibre channel arbitrated loop process described in FC-AL standard.

“Fibre channel ANSI Standard”: The standard (incorporated herein by reference in its entirety) describes the physical interface, transmission and signaling protocol of a high performance serial link for support of other high level protocols associated with IPI, SCSI, IP, ATM and others.

“FC-1”: Fibre channel transmission protocol, which includes serial encoding, decoding and error control.

“FC-2”: Fibre channel signaling protocol that includes frame structure and byte sequences.

“FC-3”: Defines a set of fibre channel services that are common across plural ports of a node.

“FC-4”: Provides mapping between lower levels of fibre channel, IPI and SCSI command sets, HIPPI data framing, IP and other upper level protocols.

“LIP”: Loop initialization protocol primitive.

“L_Port”: A port that contains Arbitrated Loop functions associated with the Arbitrated Loop topology.

“SES”: SCSI Enclosure Services.

“TPE”: Transmission Protocol Engine, a controller that operates at the FC-1 level.

To facilitate an understanding of the preferred embodiment, the general architecture and operation of a fibre channel system will be described. The specific architecture and operation of the preferred embodiment will then be described with reference to the general architecture of the fibre channel system.

FIG. 1 is a block diagram of afibre channel system100 implementing the methods and systems in accordance with the adaptive aspects of the present invention.System100 includes plural devices that are interconnected. Each device includes one or more ports, classified as node ports (N_Ports), fabric ports (F_Ports), and expansion ports (E_Ports). Node ports may be located in a node device,e.g. server103,disk array105 andstorage device104. Fabric ports are located in fabric devices such asswitch101 and102. Arbitratedloop106 may be operationally coupled to switch101 using arbitrated loop ports (FL_Ports).

The devices ofFIG. 1 are operationally coupled via “links” or “paths”. A path may be established between two N_ports, e.g. betweenserver103 andstorage104. A packet-switched path may be established using multiple links, e.g. an N-Port inserver103 may establish a path withdisk array105 throughswitch102.

In one aspect of the present invention, an acceleration method and system is provided that maintains FC-AL operation while reducing the overall latency as compared to standard FC-AL implementations.

FIG. 4 is a block diagram of an 18-port ASIC FC element400A (also referred to as system307) according to one aspect of the present invention. FC element400A provides various functionality in an FC-AL environment, including without limitation, FC element400A operates as a loop controller and loop switch usingswitch matrix408, in accordance with the FC-AL standard.

FC element307 of the present invention is presently implemented as a single CMOS ASIC, and for this reason the term “FC element” and ASIC are used interchangeably to refer to the preferred embodiments in this specification. AlthoughFIG. 4 shows 18 ports, the present invention is not limited to any particular number of ports.

System307 provides a set of port control functions, status indications, and statistics counters for monitoring the health of the loop and attached devices, diagnosing faults, and recovering from errors.

ASIC307 has 18 ports where 16 ports are shown as numeral405 while ahost port404 and cascadeport404A are shown separately for convenience only. These ports are generic to common Fibre Channel port types, for example, L_Ports.

For illustration purposes only, all ports are drawn on the same side ofASIC307 inFIG. 4. However, the ports may be located on any side ofASIC307. This does not imply any difference in port or ASIC design. Actual physical layout of the ports will depend on the physical layout of the ASIC.

Each port has transmit and receive connections to switchmatrix408 and includes transmitprotocol engine407 and a serial/deserializer406. Frames enter/leave the link405A andSERDES406 converts data into 10-bit parallel data to fibre channel characters.

Switch matrix408 dynamically establishes a connection for loop traffic.Switch matrix408 includes a global arbiter (hence switchmatrix408 is also referred to as SGA408) that provides lower latency and improved diagnostic capabilities while maintaining full Fibre Channel Arbitrated Loop (FC-AL) compliance.

Switch matrix408 is also referred to as anacceleration device408 throughout this specification. All transmitters and receivers of all devices are connected to theacceleration device408, as shown inFIG. 3. Any device that is not operating properly is isolated, which means thatacceleration device408 also acts as a virtual bypass element for each FCD. Each FCD connected to theacceleration device408 operates in the same fashion as in the traditional implementation, per the FC-AL standard.

FIG. 7 shows a block diagram ofacceleration device408 and it includes an AL_PA table408C, astate machine408D, aglobal arbiter408B, acontroller408F, anddata switching paths408A. Devices are coupled viamatrix408E throughconnections408A.State machine408D states are described belo v with respect toFIG. 9.

Global arbiter408B selects the arbitration winner from all arbitrating channels according to FC-AL rules. For switch channels that request arbitration at a given time, the AL_PA is looked up in table408C to validate an arbitration request. This is because a switch channel may be repeating an arbitration primitive that it received and may not be requesting arbitration for itself.

Arbiter408B has a validation and comparison section (not shown). The validation section includes combinatorial logic that validates an arbitration request (“cARBDet”) for a given switch channel. The comparison section includes comparators and multiplexers that output the highest arbitrating switch channel.

Controller408F has three modes of operation, loop mode, spray mode and connection mode. In the loop mode, an isolated switch channel receives data from a previous un-isolated switch channel, however data from the isolated switch channel (or port) is not sent to the loop.

In the spray mode, again an isolated switch channel receives spray elements from a previous un-isolated switch channel, however data from the isolated switch channel (or port) is not sent to the loop.Controller408F is in spray mode when there is no active traffic for the loop and data is sprayed from a host port to the transmit sections of other switch channels.

In the connection mode, isolated switch channels receive ARB(F0), however data from the isolated switch channel (or port) is not sent to the loop.Controller408F is in connection mode when there is active traffic for the loop.

Switch matrix408 provides a quasi-direct architecture in the form of a buffer-less Switch Matrix.Switch matrix408 includes data multiplexers that provide a path to each port. In one aspect, twenty multiplexers may be used. In one aspect, data is 16 bits wide plus the internal “K” control signal and two parity bits.

At power-up,SGA408 is setup in a flow-through configuration, which means all ports send what was received onhost port404. When a valid LIP sequence occurs,SGA408 configures the switch to a complete loop configuration for the address selection process. During normal data transfer on the loop,SGA408 reconfigures the switch data-path to connect the active ports in what appears as a smaller loop, which lowers the latency but still emulates FC-AL functionality to all entities on the loop.

During loop configuration,SGA408 configures the switch data-path to include a snooping port that walks through each port during the LIP physical address assignment to track each port's assigned arbitrated loop physical address (AL_PA). This snooping process is called the ‘LIP walk’. When the LIP process is done, the firmware records the “port to AL_PA” map in an internal table built inSGA408. During normal data transfer mode,SGA408 monitors arbitration requests, open requests, and close primitives to determine which ports have traffic that must be forwarded. The ports that have traffic for the loop provide the necessary information to create the connection points for the switch data-path. The inactive ports are provided the primitive ARB(F0).

SGA408 selects the arbitration winner, from all the arbitrating ports, according to Fibre Channel Arbitrated Loop (FC-AL) rules. For ports, which detect arbitration, the AL_PA is looked up in a Port Address Table to see if the arbitration request is valid for that port. Due to the unique purpose of port0 (Host port404),port0 never needs to win arbitration, but can detect that the arbitration winner is outside a range, or with ARB(F0)/IDLE show that devices outsidesystem307 are not arbitrating at a given time.

For arbitration detect onhost port404 to be considered valid the “ArbPS0” cannot match an AL_PA in the internal Port Address Table, which means that the source address is not in aparticular system307. For ports seeking a valid arbitration, the AL_PA determines which arbitrating device has highest priority; and typically, the port with the lowest AL_PA value is always selected as the winner.

SGA408 creates a direct ioop connection between source and destination devices. This connection methodology avoids the delay associated with data having to pass from one member of the loop to the next until the data has completed traversing the loop. In one aspect, the following formula evaluates performance of a loop connection:
Latency (word times)=n*(2*8)+disk+host=16n+12

Where n is the number ofsystems307 that comprise the FC loop, 6 is the latency of the disk drive that is part of the loop connection and 6 is typically the latency of the attached host.

System307 includes plural I2C (I2C standard compliant) interfaces412-413 that allowsystem307 to couple to plural I2C ports each having a master and slave capability.

System307 also includes a general-purpose input/output interface (“GPIO”)415. This allows information fromsystem307 to be analyzed by any device that can useGPIO415. Control/Status information419 can be sent or received throughmodule415.Timer module411 is provided to control various timer operations.

System307 also includes aSPI module414 that is used for parallel to serial and serial to parallel transfer betweenprocessor400 firmware andflash memory421 in the standard Little Endian format.

System307 also includes a Universal Asynchronous Receiver/Transmitter (“UART”)interface418 that converts serial data to parallel data (for example, from a peripheral device modem or data set) and vice-versa (data received from processor400) complying industry standard requirements.

System307 can also process tachometer inputs (received from a fan, not shown) using module417.Processor400 can read the tachometer input via a tachometer rate register and status register (not shown).

System307 provides pulse width modulator (“PWM”) outputs viamodule416.Processor400 can program plural outputs.

System307 also includes twoframe manager modules402 and403 that are similar in structure.Processor400 can access runtime code frommemory420 and input/output instructions from read onlymemory409.

Module402 (also referred to as the “diag module402”) is a diagnostic module used to transfer diagnostic information between a FC-AL and the firmware ofsystem307.

Diag module402 is functionally coupled to storage media (via ports405) via dedicated pathsoutside switch matrix408 so that its connection does not disrupt the overall loop.Diag module402 is used for AL_PA capture during LIP propagation, drive(s) (coupled to ports405) diagnostics and frame capture.

Module403 (also referred to as “SES module403”) complies with the SES standard and is functionally coupled tohost port404 and its output is routed throughswitch matrix408.SES module403 is used for in-band management services using the standard SES protocol.

When not bypassed,modules402 and403 receive primitives, primitive sequences, and frames. Based on the received traffic and the requests from firmware,modules402 and403 maintain loop port state machine (LPSM) (615,FIG. 6B) in the correct state per the FC-AL standard specification, and also maintain the current fill word.

Based on acurrent LPSM615 state (OPEN or OPENED State),modules402 and403 receive frames, pass the frame onto a buffer, and alert firmware that a frame has been received.Module402 and403 follow FC-AL buffer to buffer credit requirements.

Firmware may requestmodules402 and403 to automatically append SOF and EOF to the outgoing frame, and to automatically calculate the outgoing frame's CRC usingCRC generator612.Modules402 and403 can receive any class of frames and firmware may request to send eitherfibre channel Class 2 orClass 3 frames.

Port Management Interface (PMIF)401 allowsprocessor400 access to various port level registers,SerDes modules406 and TPE Management Interfaces509 (FIG. 5).PMIF401 contains a set of global control and status registers, receive and transmit test buffers, and three Serial Control Interface (SCIF) controllers (not shown) for accessingSerDes406 registers.

FIGS. 6A and 6B show block diagrams formodule402 and403. It is noteworthy that the structure inFIGS. 6A and 6B can be used for bothmodules402 and403.FIG. 6B is the internal data path of aFC port601 coupled tomodules402/403.

Modules402 and403 interface withprocessor400 via aninterface606. Incoming frames tomodules402 and403 are received from port601 (which could be any of theports404,404A and405) and stored inframe buffer607. Outgoing frames are also stored inframe buffer607.Modules402 and403 have a receive side memory buffer based on “first-in, first-out” principle RX_FIFO, (“FIFO”)603 and transmitside TX_FIFO FIFO604 interfacing with aRandom access FIFO605. A receiveside FIFO603 signals to firmware when incoming frame(s) are received. A transmitside FIFO604 signals to hardware when outgoing frames(s) are ready for transmission. Aframe buffer607 is used to stage outgoing frames and to store incoming frames.Modules602 and602A are used to manage frame traffic fromport601 tobuffers603 and604, respectively.

Modules402 and403 use various general-purpose registers608 for managing control, status and timing information.

Based on the AL_PA,modules402 and403 monitor received frames and if a frame is received for a particular module (402 or403), it will pass the frame onto a receive buffer and alert the firmware that a frame has been received via a receiveside FIFO603.Modules402 and403 follow the FC-AL buffer-to-buffer creditrequirements using module616.Modules402 and403 transmit primitives and frames based on FC-AL rules. On request,modules402 and403 may automatically generate SOF and EOF during frame transmission (using module613). On request,modules402 and403 may also automatically calculate the Cyclic Redundancy Code (CRC) during frame transmission, usingmodule612.

Overall transmission control is performed bymodule611 that receives data, SOF, EOF and CRC. Transmit Buffer control is performed bymodule614. Aword assembler module609 is used to assemble incoming words, and a fill word module610 receives data “words” before sending it tomodule611 for transmission.

FIG. 5 shows a block diagram of the transmission protocol engine (“TPE”)407.TPE407 maintains plural counters/registers to interact with drives coupled toports405. EachTPE407 interacts withprocessor400 viaport manager interface401.

Each Fibre Channel port of system400A includes a TPE module for interfacing withSerDes406.TPE407 handles most of the FC-1 layer (transmission protocol) functions, including 10B receive character alignment, 8B/10B encode/decode, 32-bit receive word synchronization, and elasticity buffer management for word re-timing and TX/RX frequency compensation.

SerDes modules406 handle the FC-1 serialization and de-serialization functions. EachSerDes406 port consists of an independent transmit and receive node.

TPE407 has a receive module500 (that operates in the Rx clock domain503) and a transmitmodule501.Data502 is received fromSERDES406 and decoded by decodingmodule504. Aparity generator module505 generates parity data.SGA interface508 allows TPE to communicate with switch514 orswitch matrix408. Interface508 (via multiplexer507) receives information from areceiver module506 that receives decoded data fromdecode module504 and parity data frommodule505.

Management interfaces module509 interfaces withprocessor400. Transmitmodule501 includes aparity checker511, atransmitter510 and anencoder512 that encodes 8-bit data into 10-bit data. 10-bit transmit data is sent to SERDES406 viamultiplexer513.

Port Management Interface (PMIF)401 allowsprocessor400 access to various port level registers,SerDes modules406 and TPE Management Interfaces509 (MIFs).PMIF401 contains a set of global control and status registers, receive and transmit test buffers, and three Serial Control Interface (SCIF) controllers (not shown) for accessingSerDes406 registers.

The following provides a description of various standard FC-AL processes that are accelerated, using system307 (and acceleration device408), according to one aspect of the present invention (also referred to as the “accelerated implementation”).

LIP Process: During the LIP process,acceleration device408 connects all devices as if they were in a loop and the devices behave as in the traditional implementation. In one aspect of the present invention,acceleration device408 monitors the frames sent during the LIP process to determine the AL_PAs selected by each FCD. A FC-AL compliant entity advances through the loop to intercept the initialization frames from each device for processing before passing them on to the next device. This retrieves the AL_PA selected by each device and creates a table for the AL_PAs of the FCDs connected tosystem307. The AL-PA table408C is then used for identification purposes for port selection.

Arbitration process: In the accelerated implementation (i.e. with acceleration device408) when the loop is idle, “host”port404 sends data to all the connected FCDs. When only one device wants access to the loop, theglobal arbiter408B connects “host”port404 to the FCD arbitrating and then connects this FCD back to the “host”port404. The other FCDs receive a benign primitive, so if they desire to access the loop they can start the arbitration process just as they would in a traditional implementation.

Global arbiter408B follows the FC-AL rules for arbitration and changes the winning port if a higher priority port starts arbitrating before any existing arbiter has won arbitration, just as in the traditional implementation.

In the accelerated implementation, only the arbitrating and “host”port404 are connected to the loop. Thus, the delay to retrieve the arbitration primitive (win arbitration) is two re-timing delays, one in theacceleration device408 and the other through the FCD on thehost port404. If the device onhost port404 is the arbitrating device then the delay is one re-timing delay through the acceleration device.

Open process: In the accelerated implementation (i.e. with acceleration device408) when the winning FCD sends out an open primitive, theacceleration device408 connects the destination FCD to the winning FCD, looping throughhost port404, and the open primitive is directly passed to the destination FCD. The destination FCD can then send the R_RDY primitive, which travels through theacceleration device408 and the FCD on thehost port404. The other FCDs receive a benign primitive to maintain the proper LPSM state in those devices. In this case, the latency is three re-timing delays, due to the open primitive, the R_RDY primitive and the delay of the FCD on thehost port404.

If the destination port is upstream of the winning port, there are three re-timing delays, two due to the open primitive and one due to the R_RDY primitive. However, if one of the ports ishost port404, then there is only two re-timing delays through theacceleration device408.

Open replicate process: In the accelerated implementation (i.e. using acceleration device408) to complete the loop circuit, the winning FCD can send an open replicate primitive(s). The winning FCD can do a broadcast open replicate or a specific open replicate and theacceleration device408 responds in a similar manner. When theacceleration device408 sees the open replicate, it creates a spray pattern that includes the FCDs ports that a source (the winning FCD) port wants to communicate with. The frames are sent to all ports at the same time and then returned to the source for removal from the loop, as per the FC-AL standard. Latency is reduced by the frames arriving at all ports at the same time instead of traveling around the entire loop.

Close process: In the accelerated implementation (i.e. using acceleration device408) since only the active devices are connected there is only re-timing andhost port404 delay to close the circuit.

State Machine: In the accelerated implementation (i.e. using acceleration device408), all FCDs connected to theacceleration device408 need a LPSM as in the traditional implementation. However,acceleration device408, only requires a state machine (408D) that describes the state of the loop, not individual loop ports. This simplified state machine is not a part of the current FC-AL standard.FIG. 9 (i.e.FIGS. 9A and 9B) shows a state machine diagram foracceleration device408, according to one aspect of the present invention.

FIG. 9 shows eight basic states forstate machine408D. Table I below describes the basic eight states:

	State	Description
	PWRUP	The power up state is thehardware reset
	State
0	state.TPE 407 uses this time for speed
		negotiation, before it allows a valid LIP to
		be indicated to theSGA 408. The only way to
		exit this state is to have both the Switch
		Enable bit set and a valid LIP indication.
		The Switch Enable bit is only evaluated in
		this state and is set by firmware ofsystem
		307.
	LOINIT	This state is used to put the switch (307) in
	State 1	the loop configuration and to process LISM
		and Lix A frames. The first switch channel to
		indicate ARB (F0) during the LISM processing
		is the Loop Initialization Master (LIM). The
		Lix A frames are snooped at each switch
		channel by walking theDiagnostics port 402
		through all unisolated switch channels.
		Firmware is responsible for advancing the
		Diagnostics port 402.
	IDLE	This state is used when there is notraffic
	State
2	on the loop. The switch (307) is set in a
		single spray configuration where the source
		of the spray is theHost Port 404.
	ARBING	This state is used to connect thehighest
	State
3	arbitrating switch channel to the loop and
		stays here until the switch channel
		arbitrating recognizes that it has won
		arbitration.
	ARB WON	This state is used to wait for anopen
	State
4	primitive once an arbitrating switch channel
		has declared it has won arbitration.
	OPEN	This state indicates that an OPEN (OPN)
	State 5	connection exists.
	CLEAN	This state is used to clear up all
	State 6	retransmitted closes after Replicated Open.
	SNDCLS	This state is used to send a closeprimitive
	State
7	for a connection switch channel that was
		isolated before it sent a Close (CLS)
		primitive itself.
	WTCLS	This state is used to wait out thecompletion
	State
8	of a connection where one of the connection
		switch channels was isolated.

Table II below shows thevarious State Machine408D transitions, according to one aspect of the present invention:

Current
State	Next State	Reason for Transition
PWRUP	PWRUP	When reset occurs, thestate
		machine
408D stays in this state
		to indicate to theSwitch
		Controller
408F that it should be
		in default single spray mode. To
		leave this state is to have a valid
		LIP occur with the Switch Enable
		bit set. Loop traffic should NOT be
		considered reliable until LIPs
		occur.
PWRUP	LOINIT	Whenswitch 307 is enabled and a
		valid LIP occurs, thestate machine
		408 changes to the loop
		initialization state. Loop traffic
		at this instance is LIP primitives.
Any	LOINIT	When a valid LIP occurs in any
State		state, thestate machine 408D
		transitions to the loop
		initialization state and loop
		traffic is LIP primitives.
LOINIT	LOINIT	State machine	408D stays in this
		state until the LIP process has
		completed. The LOINIT state
		indicates to theSwitch Controller
		408F that it should be in loop
		mode. In loop mode, firmware
		advancesport 402 for snooping
		AL_PA assignments. Loop traffic
		consists of LISM and LIxA frames.
LOINIT	IDLE	This transition occurs when the CLS
		primitive for the LIP process is
		seen. Only one CLS primitive is
		used for the LIP process. Loop
		traffic is CLS and idle primitives.
IDLE	IDLE	When none of the switch channels
		require active connections to the
		loop, thestate machine 408D stays
		in the idle state. Loop traffic is
		idle and possibly link primitives.
IDLE	ARBING	Whenever any switch channel
		indicates that arbitration
		primitive has been seen, thestate
		machine
408D changes to the
		arbitrating state. TheSwitch
		Controller
408F makes the winning
		switch channel, according to the
		Arbiter module 408B, the source
		connection for loop traffic and the
		loop traffic is arbitration
		primitives.
IDLE	OPEN	Whenever an open primitive has been
		seen on a previous source port, the
		state machine 408D changes to the
		open state. TheSwitch Controller
		408F makes the source of the open
		primitive to be the current source
		and the destination of the open
		primitive to be the destination
		connection point. Loop traffic is
		open and idle primitives.
ARBING	ARBING	While a winner is being selected,
		according to FC-AL protocol a port
		receives back it's own arbitration
		primitive to win arbitration,
		hence, thestate machine 408D stays
		in the arbitrating state. The SrcPN
		primitive can change during this
		state and loop traffic is
		arbitration primitives.
ARBING	ARBWON	When the source switch channel
		issues an ARB(F0) primitive the
		state machine 408D changes to the
		arbitration won state. Loop traffic
		is arbitration primitives.
ARBING	OPEN	When an open primitive is seen on
		the source switch channel, the
		state machine 408D changes to the
		open state. TheSwitch Controller
		408F makes the destination of the
		OPN as the destination connection
		for the loop. Loop traffic is OPN,
		arbitration and idle primitives.
ARBING	IDLE	When the source switch channel is
		isolated or when a close connection
		is seen and there are no other
		switch channels arbitrating, the
		state machine 408D returns to the
		idle state.
ARBWON	ARBWON	Waiting for an open primitive from
		the source connection, thestate
		machine
408D stays in the
		arbitration won state.
ARBWON	OPEN	When an open primitive is seen on
		the source switch channel, the
		state machine 408D changes to the
		open state. TheSwitch Controller
		408F makes the destination of the
		OPN as the destination connection
		for the loop. Loop traffic is OPN,
		arbitration and idle primitives.
ARBWON	ARBING	When a close connection is seen and
		there are other switch channels
		arbitrating, thestate machine 408D
		changes to the arbitrating state.
ARBWON	IDLE	When the source switch channel is
		isolated or when a close connection
		is seen and there are no other
		switch channels arbitrating, the
		state machine 408D returns to the
		idle state.
OPEN	OPEN	While there is an open connection,
		thestate machine 408D stays in the
		open state. Loop traffic is frames,
		R_RDYs, idle or arbitration
		primitives.
OPEN	ARBING	When a close connection is seen and
		there are other switch channels
		arbitrating and the previous
		connection was not a replicate
		spray, thestate machine 408D
		changes to the arbitrating state.
		Loop traffic is arbitration
		primitives.
OPEN	IDLE	When a close connection is seen and
		there are no other switch channels
		arbitrating and the previous
		connection was not a replicate
		spray, thestate machine 408D
		changes to the idle state. Loop
		traffic is idle primitives.
OPEN	CLEAN	When a close connection is seen and
		the previous connection was a
		replicate spray, thestate machine
		408D changes to the clean state.
		Loop traffic is idle or arbitration
		primitives.
OPEN	SNDCLS	When one of the switch channels is
		isolated and the CLS was not seen
		for that connection point, the
		state machine 408D changes to the
		send close state (“SNDCLS”). Loop
		traffic is frames and idle or
		arbitration primitives.
OPEN	WTCLS	When one of the connection switch
		channels is isolated and the close
		was seen for that connection point,
		thestate machine 408D changes to
		the wait close state (“WTCLS”).
		Loop traffic is frames and idle or
		arbitration primitives.
CLEAN	CLEAN	While the switch channels that were
		part of the replicate spray are
		retransmitting the close primitive,
		thestate machine 408D stays in the
		clean state. Loop traffic is idle
		or arbitration primitives.
CLEAN	ARBING	When the switch channels that were
		part of the replicate spray have
		completed retransmitting the close
		primitive and there is another
		switch channel arbitrating, the
		state machine 408D changes to the
		arbitrating state. Loop traffic is
		arbitration primitives.
CLEAN	IDLE	When the switch channels that were
		part of the replicate spray have
		completed retransmitting the close
		primitive and there is no other
		switch channel arbitrating, the
		state machine 408D changes to the
		idle state. Loop traffic is idle
		primitives.
CLEAN	OPEN	When the switch channels that were
		part of the replicate spray have
		completed retransmitting the close
		primitive and there is an open
		primitive, thestate machine 408D
		changes to the open state. Loop
		traffic is open, arbitration or
		idle primitives.
SNDCLS	SNDCLS	Until the correct phase of
		MsHalfWord and cDataEn occur, the
		state machine 408D stays in the
		send close state. Loop traffic is
		close, idle or arbitration
		primitives or frames.
SNDCLS	WTCLS	When there is no connection timeout
		at a next clock edge on the correct
		phase of MsHalfWord and cDataEn,
		thestate machine 408D changes to
		the wait close state. Loop traffic
		is close, idle or arbitration
		primitives or frames.
SNDCLS	IDLE	When there is a connection timeout
		at the next clock edge on the
		correct phase of MsHalfWord and
		cDataEn, thestate machine 408D
		changes to the idle state. Loop
		traffic is close, idle or
		arbitration primitives or frames.
WTCLS	WTCLS	While the connection is completing
		after one of the connection points
		is isolated, thestate machine 408D
		stays in the wait close state. Loop
		traffic is close, idle or
		arbitration primitives.
WTCLS	CLEAN	When there is a close connection
		and the previous connection was a
		replicated spray, thestate machine
		408D changes to the clean state.
		Loop traffic is close, idle or
		arbitration primitives.
WTCLS	ARBING	When there is a close connection
		and there is another switch channel
		arbitrating, thestate machine 408D
		changes to the arbitrating state.
		Loop traffic is close, and
		arbitration primitives.
WTCLS	IDLE	When there is a close connection
		and there is no other switch
		channel arbitrating, thestate
		machine
408D changes to the idle
		state. Loop traffic is close, and
		idle primitives.
Any	IDLE	When a connection timeout timer
state		expires, thestate machine 408D
except		changes to the idle state. The
PWRUP		timeout indicates that the switch
or		may be out of sync with the loop
LOINIT		traffic or that an active switch
		channel did not follow protocol and
		needs to be removed as an active
		connection. Loop traffic could be
		anything at this time.

Reducing Latency in the “Small Example” Case:

The following illustrates how the present invention minimizes latency in a “small example” case, i.e., where the number of FCD is small, as shown inFIGS. 2 and 3. InFIG. 3 the fibre channel devices are numbered as301-310 for clarification only, since the fibre channel devices are the same as inFIG. 2. Referring toFIGS. 2 and 3, ifFCD #1 wants to send frames toFCD #7, thenFCD #1 creates the loop circuit, sends the frames and closes the loop circuit. The following calculations show the difference in latency between traditional and the accelerated implementation. Table III below also provides a comparison between the traditional implementation and the accelerated implementation, according to the present invention:

Arbitration process: In the traditional implementation, the arbitration primitive (“ARB”) passes through all devices on the loop. In the above example,FCD #1 sends the arbitration primitive toFCD #2, which processes this primitive and sends it toFCD #3 and so on until the primitive is sent toFCD #1. The cumulative latency is based on the latency for all devices except #1. If the latency is the same for each device, such as the maximum latency in monitoring mode per the FC-AL standard then the “Total Latency” (latency)(n-1) where n is the number of devices connected to the loop. Thus in the above example Total Latency=(6 transmission word times)(10-1)=54 transmission word times, “word” is a unit for frame (or date) transmission. It is noteworthy that the various adaptive aspects of the present invention are not limited to any particular unit for frame/date transmission.

In the accelerated implementation, according to the present invention, the arbitration primitive is sent to theacceleration device408 and then looped around. If the arbitrating device is not onhost port404 it takes two re-timing delays to pass fromFCD #1 back toFCD #1. If the arbitrating device is on thehost port404, it takes one re-timing delay to pass fromFCD #1 back toFCD #1. Hence according to the present invention, the latency will be 16 and 8 transmission word times, respectively, instead of 54 transmission word times in the traditional implementation.

Open process: In the traditional implementation, the open primitive passes through all devices betweenFCD #1 andFCD #7.

Thus Total Latency=(latency)(m) where m is the number of devices between the two active devices.

In the above example there are 5 devices between the two active devices (FCD #2, #3, #4, #5, #6) and hence, the Total Latency=(6 transmission word times)(5)=30 transmission word times. Thereafter,FCD #7 sends an R_RDY primitive back toFCD #1, which has the same equation for Total Latency.

In the above example there are 3 devices between the two active devices (FCD #8, #9, #10) and the Total Latency=(6 transmission word times)(3)=18 transmission word times. Hence the total latency for the complete open process is 30+18=48 transmission word times.

In the accelerated implementation, according to the present invention, the open primitive is sent toacceleration device408 and then sent toFCD #7. The total latency is one re-timing delay to pass through theacceleration device408, which will be 8 transmission word times. ThenFCD #7 sends an R_RDY primitive back toFCD #1, which has the same Total Latency, i.e., 8 transmission word times. The total latency for complete open process is 8+8=16 transmission word times, instead of 48 transmission word times in the traditional implementation.

If neither of the active devices is onhost port404, there is another re-timing delay to complete the loop causing the Total Latency to be equal to 24 transmission word times.

Frame process: In the traditional implementation, frames pass through all devices betweenFCD #1 andFCD #7. The Total Latency=(latency)(m), where m is the number of devices between the two active devices. In the above example there are 5 devices between the two active devices (FCD #2, #3, #4, #5, #6), hence, the Total Latency=(6 transmission word times)(5)=30 transmission word times.

The frames fromFCD #7 toFCD #1 have the same equation for Total Latency. In the above example there are 3 devices between the two active devices (FCD #8, #9, #10) and the Total Latency=(6 transmission word times)(3)=18 transmission word times.

In a loop circuit created by an open replicate primitive the frame sent by the source travels the entire loop and then removed by the source device. The Total Latency=(latency)(n−1)=(6 transmission word times)(9)=54 transmission word times.

In the accelerated implementation, according to the present invention, the frames fromFCD #1 toFCD #7 are sent to theacceleration device408 and then sent toFCD #7. The total latency is one re-timing delay to pass through the acceleration device, which is 8 transmission word times. The frames fromFCD #7 toFCD #1 have the same Total Latency, i.e. 8 transmission word times. Hence, the total latency is 16 transmission word times, instead of 48 transmission word times in the traditional implementation.

In a loop circuit created by an open replicate primitive the frame sent by the source is sent to all necessary devices at the same time and then returned to the source to be removed. The total latency is two re-timing delays, which is 16 transmission word times, instead of 54 transmission word times in the traditional implementation.

If one of the active devices is not onhost port404 another re-timing delay is added, thus total latency=24 transmission word times.

Close process: The closing of a loop circuit is similar to opening a loop circuit, except that the two active devices can send their close primitives at the same time. In the worst case scenario, the difference in latency between the two implementations is the same as the open process. In the best case scenario, the two active devices have the same number of devices between them on both paths and both devices send their closes at the same time.

In this case for the traditional implementation the Total Latency=(latency)(n/2−1)=(6 transmission word times)(10/2−1)=24 transmission word times.

For the accelerated implementation, according to the present invention, the total latency is two or three re-timing delays (two if the host port has one of the active devices), i.e. 16 or 24 transmission word times.

TABLE III
Comparison of Latency in a Small Example Case

			Accelerated
		Accelerated	Implementation/
		Implementation/	Host NOT an
Latency Type	Traditional	Host an active port	active port

Arbitration	54	8	16
Open	48	16	24
Frame FCD #1 to	30	8	8/16
FCD #7
Frame FCD #7 to	18	8	16/8
FCD #1
Frames back and	48	16	24
forth (total)
Frame Open	54	16	24
Replicate
Close	24	16	24

Reducing Latency in the “Maximum Example” Case:

FC-AL standard allows for a maximum of 127 devices on a loop. All 127 FCDs could be connected to oneacceleration device408 orplural system307 may be used to operationally/functionally couple 127 devices. In one aspect, 17 FCDs may be coupled tosystem307 viaacceleration device408. To get a greater number of devices connected,system307 be cascaded together such that thecascade port404A of thefirst system307 is connected to thehost port404 of thesecond system307 and so on.FIG. 8 shows an example of howplural system307 may be used to couple 127FCDs (the “Maximum Case” (shown as devices801-804).

The following provides a comparison (including Table IV) between traditional and the accelerated implementation latency, according to the present invention in the Maximum Example Case.

Arbitration process: In the traditional implementation, the arbitration primitive passes through all devices on the loop. In the above example,FCD #1 sends out the arbitration primitive toFCD #2 that processes this primitive and sends toFCD #3 and so on, until the primitive is sent toFCD #1. The total latency is due to the latency in all devices except #1. If the latency is the same for each device, such as the maximum latency in monitoring mode per the FC-AL standard then, Total Latency=(latency)(n−1) where n is the number of devices connected to the loop. Thus in the above example Total Latency=(6 transmission word times)(127−1)=756 transmission word times.

In the accelerated implementation, the arbitration primitive is sent to theacceleration devices408 and then looped around. In one aspect, the arbitration primitive is sent to each acceleration device and then looped back. IfFCD #1 sends the arbitration primitive then there will be one retiming delay to go throughcascade port404A and then so on through allacceleration devices408 except the last, which will just take one retiming delay to loop the arbitration primitive back throughhost port404. There is a re-timing delay to get the arbitration primitive from thecascade port404A to hostport404 and so on throughplural system307. Total Latency=(2)(latency)(n−1)+latency=(2)(8 transmission word times)(8)+(8 transmission word times)=136 transmission word times, instead of the traditional 756 transmission word times.

If neither of the active devices is onhost port404 there is another retiming delay to complete the loop causing the Total Latency to be equal to 144 transmission word times.

Open process: In the traditional implementation, the open primitive passes through all devices betweenFCD #1 andFCD #7. Thus Total Latency=(latency)(m) where m is the number of devices between the two active devices. In the above example there are 5 devices between the two active devices (FCD #2, #3, #4, #5, #6) and the Total Latency=(6 transmission word times)(5)=30 transmission word times. ThenFCD #7 will send an R_RDY primitive back toFCD #1, which has the same equation for Total Latency. In the above example there are 120 devices between the two active devices (FCD #8, #9, #10 . . . #127) and the Total Latency=(6 transmission word times)(120)=720 transmission word times. The complete open process is 30+720=750 transmission word times.

In the accelerated implementation, the open primitive is sent to theacceleration device408 and then sent toFCD #7. The total latency is one retiming delay to pass through theacceleration device408. This would be 8 transmission word times. ThenFCD #7 will send an R_RDY primitive back toFCD #1, which has the Total Latency required to pass through allacceleration devices408 and loop back toFCD #1, which is 136 transmission word times. The complete open process takes 8+136=144 transmission word times, instead of 750 transmission word times in the traditional implementation.

If neither of the active devices is onhost port404 there is another retiming delay to complete the loop causing the Total Latency to be equal to 152 transmission word times.

Frame process: In the traditional implementation, the frames pass through all devices betweenFCD #1 andFCD #7. Thus Total Latency=(latency)(m) where m is the number of devices in between the two active devices. In the above example there are 5 devices between the two active devices (FCD #2, #3, #4, #5, #6) and the Total Latency=(6 transmission word times)(5)=30 transmission word times. The frames fromFCD #7 toFCD #1 have the same equation for Total Latency. In the above example there are 120 devices between the two active devices (FCD #8, #9, #10 . . . #127) and the Total Latency=(6 transmission word times)(120)=720 transmission word times.

In a loop circuit created by an open replicate primitive the frame sent by the source will travel the entire loop and then be removed by the source device. The Total Latency=(latency)(n-1)=(6 transmission word times)(126)=756 transmission word times.

In the accelerated implementation, the frames are sent toacceleration device408 and then sent toFCD #7. The total latency is one retiming delay to pass through the acceleration device. This would be 8 transmission word times. The frames fromFCD #7 toFCD #1 have the Total Latency of passing through each acceleration device and then looping back, which is 136 transmission word times, according to one aspect of the present invention, instead of 720 transmission word times in the traditional implementation.

In a loop circuit created by an open replicate primitive, the frame sent by the source is sent to all necessary devices at the same time and then returned to the source to be removed. The total latency is one extra retiming delay, which is 144 transmission word times, according to the present invention, instead of 756 transmission word times in the traditional implementation.

If one of the active devices is not on the host port another retiming delay is added, thus total latency=152 transmission word times.

Close process: The closing of a loop circuit is similar to opening a loop circuit, except the two active devices can send their close primitives at the same time. In the worst case scenario, the difference in latency between the two implementations is the same as the open process.

In the best case scenario, the two active devices have the same number of devices between them on both paths and both devices send their “closes” at the same time. In this case for the traditional implementation the Total Latency=(latency)(n/2-1)=(6 transmission word times)(127/2-1)=378 transmission word times.

For the accelerated implementation the total latency is half the total latency of the acceleration devices, 144/2=72 transmission word times, when both devices are sending close at the same time and the devices are half way around the loop from each other.

TABLE IV
Comparison of Latency in the Maximum
Example Case

			Accelerated
		Accelerated	Implementation/
		Implementation/	Host
	Traditional	Host an active	NOT an
Latency Type	implementation	port	active port

Arbitration	756	136	144
Open	750	144	152
Frame FCD #1	30	8	16
toFCD #7
Frame FCD #7	720	136	144
toFCD #1
Frames back	750	144	152
and forth
(total)
Frame Open	756	152	160
Replicate
Close	378	72	80

Although the present invention has been described with reference to specific embodiments, these embodiments are illustrative only and not limiting. Many other applications and embodiments of the present invention will be apparent in light of this disclosure and the following claims.

Claims (13)

What is claimed is:

1. A fibre channel switch element, comprising:

an acceleration device that (a) connects a plurality of fibre channel devices in an arbitrated loop; (b) monitors frames sent by of the plurality of fibre channel devices during a loop initialization process to determine an arbitrated loop physical address (AL_PA) value selected by each of the plurality of fibre channel devices; (c) creates an AL_PA table based on the AL_PA values selected by each of the plurality of fibre channel devices, wherein the AL_PA table is stored in the acceleration device and the AL_PA table is used for port selection; and (d) during an arbitration process only an arbitrating fibre channel device and a host port for the acceleration device are connected to the arbitrated loop and the acceleration device sends a benign primitive to non-arbitrating devices so that the non-arbitrating devices start the arbitration process when ports of the non-arbitrating devices are ready for arbitration;

a global arbitration module that validates arbitration requests received from the plurality of fibre channel devices by using the AL_PA table;

a controller module that operates in (i) a loop mode during which an isolated port receives data from an un-isolated port but data from the isolated port is not sent to the arbitrated loop; (ii) a spray mode when there is no active traffic for the arbitrated loop and data is sent from the host port to other ports; and (iii) a connection mode when there is traffic for the arbitrated loop and an isolated port receives a primitive but is unable to send frames to the arbitrated loop; and

a state machine that uses a state of the arbitrated loop for changing fibre channel switch element states, instead of using individual port states of the plurality of fibre channel device ports.

2. The switch element ofclaim 1, wherein during an open process, the acceleration device connects a source fibre channel device that wins arbitration to a destination device via the host port and an open primitive is sent directly from the source fibre channel device to the destination device and in response the destination device directly sends a response primitive (R_RDY) to the source fibre channel device.

3. The switch element ofclaim 2, wherein non-destination devices receive a benign primitive during the open process to maintain proper loop state machine state instead of receiving an open primitive from the source fibre channel device.

4. The switch element ofclaim 1, wherein during an open replicate process the acceleration device receives an open replicate primitive from a source fibre channel device and creates a spray pattern to send the open replicate primitive at the same time directly to all ports with which the source fibre channel device intends to communicate.

5. The switch element ofclaim 1, wherein the acceleration device after the loop initialization process and during data transfer mode reconfigures a data path to directly connect active ports.

6. The switch element ofclaim 1, wherein during a data transfer mode after the loop initialization primitive process the acceleration device monitors arbitration requests, open requests and close primitives to determine which ports have traffic for creating connection points for a data path.

7. An acceleration device of a fibre channel switch element for connecting a plurality of fibre channel devices in an arbitrated loop via a host port, comprising:

a global arbitration module that validates arbitration requests received from the plurality of fibre channel devices by using an arbitrated loop physical address (AL_PA) table populated with AL_PA values selected by the plurality of fibre channel devices and the AL_PA values are determined by the acceleration device during a loop initialization process when the acceleration device monitors frames that are sent by the plurality of fibre channel devices;

a state machine module that uses a state of the arbitrated loop for changing fibre channel switch element states, instead of using individual port states of the plurality of fibre channel device ports;

a switch matrix for connecting the plurality of fibre channel devices; and

a controller module that operates in (i) a loop mode during which an isolated port receives data from an un-isolated port but data from the isolated port is not sent to the arbitrated loop; (ii) a spray mode when there is no active traffic for the arbitrated loop and data is sent from the host port to other ports; and (iii) a connection mode when there is traffic for the arbitrated loop and an isolated port receives a primitive but is unable to send flames to the arbitrated loop; wherein the AL_PA table is used for port selection; and during an arbitration process only an arbitrating fibre channel device and the host port for the acceleration device are connected to the arbitrated loop and the acceleration device sends a benign primitive to non-arbitrating devices so that the non-arbitrating devices start the arbitration process when ports of the non-arbitrating devices are ready for arbitration.

8. The acceleration device ofclaim 7, wherein during an open process, the acceleration device connects a source fibre channel device that wins arbitration to a destination device via the host port and an open primitive is sent directly from the source fibre channel device to the destination device and in response, the destination device directly sends a response primitive (R_RDY) to the source fibre channel device.

9. The acceleration device ofclaim 8, wherein non-destination devices receive a benign primitive during the open process to maintain proper loop state machine state instead of receiving an open primitive from the source fibre channel device.

10. The acceleration device ofclaim 7, wherein during an open replicate process the acceleration device receives an open replicate primitive from a source fibre channel device and creates a spray pattern to send the open replicate primitive at the same time directly to all ports with which the source fibre channel device intends to communicate.

11. The acceleration device ofclaim 7, wherein during a data transfer mode after the loop initialization process the acceleration device monitors arbitration requests, open requests and close primitives to determine which ports have traffic for creating connection points for a data path.

12. A method for accelerating traffic in a fibre channel arbitrated loop topology where a fibre channel switch element with an acceleration device couples a plurality of fibre channel devices to the arbitrated loop via a host port, comprising:

monitoring fibre channel frames during a loop initialization process;

creating an arbitrated loop physical address (AL_PA) table based on AL_PA values selected by the plurality of fibre channel devices; wherein the AL_PA table values are used by a global arbitration module for validating arbitration requests during an arbitration process when only an arbitrating fibre channel device and the host port for the acceleration device are connected to the arbitrated loop and the acceleration device sends a benign primitive to non-arbitration devices so that the non-arbitrating devices start the arbitration process when ports of the non-arbitrating devices are ready for arbitration;

using the AL_PA table for identifying ports;

connecting a source fibre channel device that sends an OPEN primitive to a destination port;

sending a benign primitive to a non-destination device connected to the arbitrated loop so that the non-destination device maintains a proper loop state machine state;

creating a spray pattern after receiving an open replicate primitive from a fibre channel device; and

sending frames directly to a plurality of destination ports at the same time.

13. The method ofclaim 12, wherein the acceleration device includes a state machine that uses a state of the arbitrated loop for changing fibre channel switch element states, instead of using individual port states of the plurality of fibre channel device ports.

Images